ken@gvax.cs.cornell.edu (Ken Birman) (11/28/89)
I received an interesting note today, to the effect that the MIPS stack alignment fix from a few days ago also "seems to have fixed the tendency of ISIS to slowly leak memory and eventually crash". Meanwhile, someone separately asked if I had a fix to the "tendency of the ISIS protos process to leak memory and eventually crash". I was actually unaware that ISIS V1.3 has a memory leak. It would be nice to learn that this fix corrected it, but I would like some help in gathering data. At Cornell, ISIS definitely is NOT leaking core, but perhaps we just aren't exercising the right path through the system. So, if anyone observes a version of ISIS that seems to be leaking memory, please help me track down the problem. Do this by sending a "kill -USR2" to the process in question and mailing me the contents of the log file this produces. Let me know enough about your system so that I figure out what I am seeing (machine type, version of OS if known, CC or GCC, etc). The part of the log file that interests me most is the few lines that talk about memory and message use. Also, let me know what the system has been doing, if anything. Does anyone know of a way to force the system to leak rapidly and crash? Ken